Capturing Sexual Behavioral Preference to Assess Sexual Risk of Gay and Bisexual Men: a Question of Measures

Abstract:
Background/Purpose The most commonly used measures to assess sexual risk of gay and bisexual men (GBM) include capturing sexual behavioral preference with an identity label (e.g., insertive anal [AI] preference - top', receptive anal [AR] preference - bottom, preference for both AR and AI [versatile]) or measures that assess participants' actual sexual behavior. These measures are often used interchangeably, although Bauer & Jairan (2008) caution that their use may result in varying outcomes. This study examines whether using a behavioral measure of sexual preference (SB) among GBM results in different sexual risk outcomes in the same data set when using an identity measure (IL). Methods: The sample was taken from cross sectional data from a survey that used convenience sampling. Variables included demographics (age, education, race) and sexual risk (sero-discordant unprotected anal intercourse – SDUAI measured for both primary and secondary sexual partners). Variables of primary interest were self-identified IL (top, bottom, versatile) and self-reported SB (a categorical variable calculated by assessing exclusively AR and AI, and versatile anal behavior of the respondent over a 6 month period) Results: The sample included 383 participants: 40.5% were racialized men (13.8% Black, 19.6% Asian, 7.0% mixed race/other). Mean age was 34 years (SD = 12), and 75% of the participants had some post-secondary education. SDUAI was reported by 23.8%. For IL, 17.5% identified as tops, 16.7% as bottoms, and 59.5% as versatile. For SB, 18.3% reported exclusive AI, 37.6% AR and 44.1% both behaviors. Bivariate analyses showed that SL and SB were associated (χ2 (4,N=358) = 131.85, p <.001. Bivariate analysis showed sexual risk was associated with AI, (χ2 (2,N=323) = 21.13, p <.001), but not with IL (χ2 (2,N=323) = 0 .38, p = .828). An independent backward elimination logistic regression model was employed to ascertain whether there are differences between the 2 measures when used independently to explain the relationship between demographic variables and sexual risk. In model 1, IL was entered with demographics to predict sexual risk. Only race was significant - Asian identifying men were less likely to report sexual risk than their White counterparts (OR 0.39, 95% CI 0.17, 0.93 p<.05). IL did not contribute to the model and was removed in the first of 4 steps. In model 2, which used SB with demographics to predict sexual risk, race was not predictive, only SB: Those engaging in AI were more likely than those reporting AR to report risk (OR 0.23, 95% CI 0.16, 0.81 p<.01). Conclusions/Implications: AI is associated with sexual risk among GBM. And yet, in the same data set an associated measure, identity label, was not found to be associated with sexual risk. Reducing sexual risk among GBM hinges on our ability to employ precise methodological procedures to analytically capture behaviors that may compromise sexual health. This study suggests that it is critical for researchers to employ measures of both SI and SB when assessing the sexual risk of GBM in order to ensure accuracy of results.